Volume 51, Issue 5, Pages 661-670 (September 2006) The Radial Bias: A Different Slant on Visual Orientation Sensitivity in Human and Nonhuman Primates  Yuka Sasaki, Reza Rajimehr, Byoung Woo Kim, Leeland B. Ekstrom, Wim Vanduffel, Roger B.H. Tootell  Neuron  Volume 51, Issue 5, Pages 661-670 (September 2006) DOI: 10.1016/j.neuron.2006.07.021 Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 1 Psychophysical Testing Shows a Higher Contrast Sensitivity for Radial Orientations, Compared to Tangential Orientations (A) Stimulus configuration. A small grating patch was presented peripherally. In the example here, the grating orientation was radial, and it was located in the upper right visual field, in position #2. In the actual experiments, a patch of either radial or tangential orientation was presented at each of eight possible locations, indicated by dotted lines, relative to the fixation point (central cross). (B) Results. Contrast sensitivity (the inverse of contrast detection threshold) is plotted, at each of eight locations. Error bars represent one standard error of the mean. When we applied a repeated measure of ANOVA with orientation (radial and tangential) and location (eight possible locations) as factors, the test revealed a significant main effect of orientation (df = (1, 7), p < 0.0001). The interaction of orientation and location (df = (7, 24), p < 0.025) demonstrated that the sensitivity for radial orientations was significantly better than that for tangential orientations. Because the interaction indicated that the sensitivity depends on visual field position, we further analyzed that effect. The apparent sensitivity difference between the upper and lower visual fields did not reach statistical significance. Sensitivity along horizontal and vertical meridians was significantly better than that for oblique meridians (p < 0.05), consistent with the well-known oblique effect. Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 2 The Radial Orientation Hypothesis Predicts that Gratings of Orthogonal Oblique Orientation Will Selectively Activate the Retinotopic Representations of Orthogonal Visual Field Quadrants To illustrate this prediction, (A) divides the visual field into quadrants, indicated by arbitrarily assigned colors. Solid lines indicate the retinotopic representation of the horizontal meridian, dotted lines indicate the upper vertical meridian, and dashed lines indicate the lower vertical meridian. (B) A map of corresponding visual field representations in human visual cortex (flattened right and left hemisphere on the right and left respectively), acquired from a representative subject using conventional fMRI retinotopic mapping. Thus, when one stares at the center of (A), stimuli in the upper left of the visual field (red) produce activity in the lower right of the flattened cortex ([B], red), and similarly for each of the other colors. At a finer level of analysis, each of the color-coded cortical “quadrants” is comprised of four repeated maps of that same visual field quadrant. For example, the upper left (red) visual field is represented in each of four cortical areas (V1v, V2v, V3v, and V4v). However, since these four cortical areas lie adjacent to each other, the overall retinotopic organization shown in (B) remains true. The retinotopic mapping is similar in macaque monkey (e.g., Figure 5). (C) Examples of the stimuli used in the orientation tests. Pseudocolor has been added here to illustrate that only alternating visual field quadrants contain radial orientations; the actual experimental stimuli were black-white. (D) The activation predicted by the gratings in (C), with the location of the retinotopic visual areas included. Consistent with the radial orientation hypothesis and the retinotopic map, the relative activity differences produced by the orthogonal orientations should reverse in sign between hemispheres, and the radial orientation bias should extend over most or all of the retinotopic cortical areas. Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 3 As Predicted, Gratings of Orthogonal Oblique Orientation Activate Complementary Quadrants in Visual Cortex, in Human and Nonhuman Primates (A) Examples of the stimuli used in this experiment. Subjects fixated the center of both stimuli, which extended over the entire visual field representation activated in (B) and (C). As in Figure 2, color has been added here to the experimental stimuli, to clarify the relationship between the stimuli and the corresponding cortical activation (B and C). (B) The activity maps from the left and right hemispheres of the human subject whose right hemisphere is illustrated in Figure 2 (to facilitate comparison to the experimental prediction). As in the icon (top, [A]), significantly higher activity in response to the 10:30–4:30 oblique orientation (45°) is shown in red-yellow pseudocolor; higher activity to the orthogonal oblique grating is blue-cyan. As predicted, fMRI activity produced by this change in orientation is relatively higher in complementary cortical quadrants. The color bar to the right indicates the statistical significance of the fMRI activation. (C) The analogous fMRI result, produced by the same stimuli, from the analogous regions of cortex, in an awake fixating monkey. Figure conventions (and most experimental details) are as in (B). Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 4 Responses to Orthogonal Oblique Orientations, Averaged across Subjects, in fMRI and Psychophysics (A) fMRI activity in flattened left and right hemispheres of visual cortex (as in Figures 2D and 3B), in response to two orthogonal oblique orientations (e.g., Figures 2C and 3A), averaged across all subjects tested (map threshold, p < 0.001). The same biases shown in the individual subject (Figure 3B) are confirmed in this average across subjects. (B) Amplitude of MR signal changes averaged from cortical areas V1, V2, and V3, corresponding to each visual field quadrant, averaged across subjects. (C) Psychophysical data from corresponding regions of the visual field (as in Figure 1), also averaged across subjects. Note the similarity between (B) and (C), despite the difference in scale across the two dimensions. Error bars represent one standard error of the mean. Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 5 Retinotopic Confirmation of the Radial Orientation Effect in Macaque Monkey (A) Magnified view of Figure 3C, showing activity differences in response to orthogonal oblique orientations. As in the human experiments, the monkey fixated the center of the gratings, near continuously throughout the fMRI acquisitions (e.g., Figure 1 of Vanduffel et al. [2001]). To confirm that these variations in fMRI activity matched the specific predictions of the radial orientation hypothesis, we also activated the representation of the upper and lower visual fields (B) and the vertical and horizontal meridians (C), using fMRI coupled with retinotopically varying stimuli (Fize et al., 2003) in the same fixating monkey, in separate experiments. This revealed the location of the retinotopy and cortical areas in the same cortical tissue ([D] solid lines indicate the retinotopic representation of the horizontal meridian, dotted lines indicate the upper vertical meridian, and dashed lines indicate the lower vertical meridian). As predicted, the relative activity differences produced by the orthogonal orientations reversed in sign between hemispheres and along the retinotopic division between upper and lower visual fields. The radial orientation bias included essentially all the retinotopic visual areas. On-line recordings of eye movements and the abrupt functional transition in the foveal representation of the retinotopic maps (e.g., [B] and [C]), confirmed that central visual fixation remained near-constant during the fMRI acquisitions. Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 6 fMRI Reveals a Systematic Relationship between Orientation Sensitivity and the Retinotopic Map in Visual Cortex, Based on Enhanced Activity to Radial Orientations The top panels are examples of the grating stimuli used here, oriented either vertically (A) or horizontally (B). Again, color was not present in the actual experimental stimuli, but pseudocolor has been superimposed on the stimuli here, to indicate the location of radial stripes. In response to the vertical grating, any enhanced responsiveness to radial orientations will produce higher activity along the vertical retinotopic meridian. In the horizontal grating, the radial stimulus stripes (and the predicted enhancement of activity) are rotated 90°. (C–F) Topography of fMRI activity in human visual cortex from the right hemisphere, in flattened cortical maps. (C, D, and F) From three different subjects; each shows significant increases in activity produced by gratings of either vertical orientation (red-yellow) or horizontal orientation (blue-cyan) in different cortical regions. (E) Representative map of phase-encoded polar angle retinotopy, from the hemisphere also shown in (C). As in other figures, solid lines indicate the retinotopic representation of the horizontal meridian in the same subject, dotted lines indicate the upper vertical meridian, and dashed lines indicate the lower vertical meridian (logo, leftmost panel [E]). As predicted by the radial orientation hypothesis (A and B), the vertical grating produced relatively higher activity along retinotopic representations of the vertical meridian (e.g., borders of V1–V2, V3–V4v, V3–V3A). The horizontal grating produced higher activity along representations of the horizontal meridian (e.g., V2–V3 border, midline of V1). Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 7 fMRI Activity Produced by Oblique Orientations Is Significantly Higher along the Representation of Corresponding Retinotopic Polar Angles (A) Examples of the orientations tested. Superimposed on the actual stimuli are the regions of the visual field represented in cortex (B) (dashed line = left upper visual field, right inferior visual cortex; solid line = right upper visual field, left inferior visual cortex). Red, green, yellow, and blue pseudocolor highlight the location of the radial stimulus stripes in each stimulus. (B) Magnified view of maximal fMRI activity in the flattened inferior left and right hemispheres (on the left and right, respectively), produced by gratings of either vertical (red), horizontal (blue), or oblique (green = left-leaning, or yellow = right-leaning) orientation. Corresponding representations of the retinotopic meridians are indicated in the same subject's cortex as solid and dotted lines. The activity produced by each grating was subtracted against that produced by the otherwise-equivalent grating at the orthogonal orientation. Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions

Figure 8 Control Tests Clarify the Radial Orientation Bias in Human Visual Cortex These four panels show fMRI maps taken from the same subject, in the flattened inferior right hemisphere. In all panels, the stimuli and activity format are as described in Figure 6. (A) Relative activity during passive viewing conditions. (B) Analogous activity while subjects concurrently performed an attention control experiment. (C) Activity when gratings were presented in event-related format (stimulus duration = 1 s; ISI = 1–22 s). (D) Activity when averaged across all subjects, with cortical surfaces aligned according to anatomical (gyral/sulcal) landmarks (e.g., Dale et al. [1999], Fischl et al. [1999]); thus, individual retinotopic borders are not shown. Although there is some variability in results across the different scan sessions, the basic result is similar in all conditions: vertical orientations produce highest activity (red-yellow) along the vertical retinotopic meridian (dotted lines), and horizontal orientations produce highest activity (blue-cyan) along the horizontal meridian (solid lines). Neuron 2006 51, 661-670DOI: (10.1016/j.neuron.2006.07.021) Copyright © 2006 Elsevier Inc. Terms and Conditions